How to FIX AW17R4/15R3/13R3 CPU Core Temperature Differential Issue

I am remembering something, back in the days, like way back I had an AMD Athlon 1200mhz Thunderbolt CPU. Those had a bare die out of the box and a lot of people broke those CPU dies while mounting a heavy super heavy cooler on top of it. The sceramic corners would crack if the cooler was mounted slightly uneven. To counter that there where copper shims available to surround the CPU die and make sure the cooler was balanced properly.



I wonder if something like this made for a mobile core I7 would even out the heatsink as well and ensure perfect pressure onto the CPU die.

 

This would be a regression.. it's not that tough to make a normal even heat sink with 4 mounting points. Only if it would be profitable for the manufacturer. Not a large number of companies want to make lasting machines.

 

I came across the same issue with 4 mounting points around the CPU (but also the GPU mounting points and even the fan mounting points influence the stability. The CPU die is just too small in my opinion. So in all cases this would help. Unfortunately I do not have the technical knowledge to make a precision cut shim (and also not sure if the height is the same in all cases).

 

What changes are you discussing with the seller?

I'm thinking about ordering one soon.

 

I have a ranger

 

*please delete this blank reply*

 

What about this solution? Big fake or Okay ?

 

One member bought it and tested...well it is nothing special, ~3C cooler

 

So... Big fake!

 

The point is differential Temps.

You can't defy physics with an aftermarket cooler

 

Uneven pressure of the heatsink is the reason...not only on this "tripod", but also it is present on quite few "quadpod" also...long live to those brilliant engineers

 

I see many laptops on notebookcheck with uneven core temps. Gt75vr, the new omen x, even clevos with desktop cpus.

Starting to think it’s more due to intels design and thermal management, as some cores boost higher than others, makes it harder to manage.

If even one or two cores are hottest but still safely under throttling temps, and they have a 10C dif from the coolest cores does it impact performance significantly?

 

Yeah thats where my findings as well. My previous non "tripod" alienware also had it a bit. But kinda bummed out that with my current model that I had only 3C and now 5~6C after a repaste, even though the temps are about 5C lower in general during load. The problem are the surrounding components which can vary between various batches yet they all use the same thickness thermal pads on their heatsinks that do not compensate for this.

The other Alienwares that I repasted had less core temperature differences. It's just so finicky the curretn Intel setup, the cores are too small and balancing the heatsink on these generations is exceptionally difficult. Rather saw this CPU's with a proper heatspreader.

 

I'm thinking the same thing. The same issue was happening with my 6820HK in a 17R3 as well.

Did anyone investigate the claims of slightly bent silicone dies? There were some convincing pictures showing a curve in the reflections of the die.

 

I cant imagine this to be honest. CPU die's are incredibly brittle and hard, they cannot bent without breaking. Also they are very highly precision made.

 

Actually back in the old days, people *DID* sand their slugs. Yes, exposed DIE. Pentium 3 days.

 

You refer to this one? http://forum.notebookreview.com/thr...ifferential-issue.805062/page-8#post-10552641

 

Was that because there was some kind of coating on top of it? The die itself is incredibly weak. But the current CPU's dont have such a thick coating, pretty much only on the sides to prevent corners breaking off.

Looks like light refraction to me.

 

Hi. I have a 15r3 7700 1070 09/02/17
Repaste Kryonaut + PCH Mod
Radiator SUNON
https://imgur.com/Bx6SDf7
https://imgur.com/a/6BgQs
https://imgur.com/a/f1Bz0

 

Please use IMGUR for hosting pictures.
It's free, everyone can see it and there is NO cost nor strings attached, and you can make an account in 1 minute.
Then you can just upload then post the "direct link" shown. Very, very easy.

 

I've been reading this very thread and keeping up. My 13R3 doesn't have a heat sink that is labeled but had core temp differences of 8c. Took off all the thermal pads and replaced with thermal paste. (oh well I don't really care) and this has solved all my temp issues. Max GPU temp and CPU temp under load all night are 75c. 65F ambient temp

 

Really. Realllly. Normal paste or gel? It looks like I'm going to open it back up today and take some pads off.

 

I am sure it's not the proper way but after taking everyone's facts into consideration and the heatsinking variances when it comes to build quality.. I will just throw paste all over it and call it a day.

I used MX4 and put just enough on all the VRMs and Memory Modules for the GPU. I figure it'll do better than the pads anyways. When the thermal pads are not attached to the heatsink, it sits very nicely on the CPU and GPU. The Thermal paste doesn't have to make up much room and besides, the paste is thick anyways. I taped up all the places the paste might get on.

Just one persons results on their 13R3.

EDIT: I run BOINC basically all day and all night. If the Thermal pad kept the VRMs and Memory Modules cool enough then I'm sure the direct thermal paste application is an upgrade..

 

Recipe for disaster. I expect the gap to be big enough for causing pump out issues. You need incredibly viscose paste to even try it.

 

Just done the same this morning and results are much better than with pads. Used ME Andromeda for testing as my machine was hitting 95-97 during gaming.
But now temps are below 80`s for cpu & below 70 for gpu after 2-3 hours constant gaming.

 

Repeated cycling of the heating up and cooling down pulls thermal paste away between the components and thus drying out. This effect is called pump out. Also if done correctly with a decent fitting heatsink you dont need to take such drastic matters. Your temps are the same as mine with just a repaste, proper assembly and i didnt even change the thermal pads.

 

I can’t even remember at the moment how many times I was trying to fit heatsink “in right place” after repaste and no significant change in temps was reached. I’ve been getting almost 100 C from the box plus huge diff between cores (about 20 degree).
Anyway I’m not going to stick forever with paste only. I just need to survive a couple of month and then planning to open laptop again and apply proper pads and paste.



Отправлено с моего iPhone используя Tapatalk

 

What revision of the heatsink do you have? 20c is quite extreme. Mine came from a late batch so out of the box i already had barely a core temp difference.

A new heatsink might worth a shot.

 

Be sure you have full warranty the whole time you have your notebook, because most laptops don't have thermal sensors for other than the cpu and gpu cores. Hardware components like vrm's without proper cooling will decrease it's lifespan a lot faster than you think. Maybe it won't die tomorrow but the expected lifespan will be short lived.

 

This, dont risk it with experimenting by using thermal paste for the chokes and vrms.

 

The funny part is, I own a lot of gaming laptops that don't have any VRM heatsinks for the GPU or CPU. One notably being the W110ER which I argue is less power efficient and those things do get HOT but it hasn't failed in any way yet. I'm sure there is a giant argument to be had but less CPU temps due to the heatsink making better contact means more hot air out the back of the laptop and less "sticking around in the heatsink due to heatsoak" so less ambient temps inside the case where... oh right the heatsink soaks it in.

I used a very small amount so hopefully most of it doesn't squirt out the sides

 

FYI. Most part of 980m died due blown vrm’s. Not from overheating cores.

 

A lot of gaming laptops also fail quite quickly. More paste is better in this case. But I sitll do not reocmmend it. Just get a revised heatsink if you dont have it already. If it sitll gives issue, than contact Dell. You paid for a working product so you should receive one. The chance of getting a good unit is now much higher than a year back.

It sucks but it is what it is. Unfortunately this CPU gen a lot of manufacturers seem to struggle with it. It's hard to find the perfect cooled machine now. But this issue can be fixed with the right heatsink.

 

I totally agree.

 

Which heatsink gets the best contact with the CPU die? I am looking to buy a new heatsink as I'm having problems with my CCI one. And will I have to repad it when I buy it?

 

I use for laptop for BOINC and it since it sits in my office connected to 5 screens i usually don't take it to go with me. It runs full bore 24/7. I'll let everyone knows if it fails with my new paste job. Temps are currently 76C on the CPU and 78C on the GPU. Stock clocks

 

In genral the heatsinks all have the same design except different heatpipe thicknesses depending on the config of what it was made for.

They all come with pads pre installed if they are new and come from a legit sourcelBut a repad yourself first might be beneficial.

 

So what is the consensus? Which are the best heatsinks? Are the Sunon S-J-1s any good?

 

dont bother hunting down which heatsink. CCI or SUNON they are both the same. they are equally good and not good. It's the shape of the heatsink that matters and unfortunately it's all luck dependent. R4 is extremely sensitive to heatsink shape due to the weak nature of the tripod cpu arm and it's very difficult to get a perfect contact between the CPU and heatsink. even no two repaste produce the same result.

 

Recently found out that thermal pads are way too important with this.

When i repasted i always also repadded on units of my friends. Thermal grizzly minus and gelid pads mostly. Those are quite soft especially the thermal grizzly ones. I didnt with my own unit because i wasnt planning to overclock. But the slight inbalance pumps out my paste. After checking the core temperatures i could only come to the conclusion that one of the ram pads ruins the perfect balance. When i remove that pad i have equal temperatures. I had 3 to 4c max stock, could lower it with a repaste but have paste pumpout. Can go to a perfect fit without pads. Now waiting for my batch of thermal grizzly minus pads to arrive at my apartment to do a full repad.

 

I'm the new guy here and I only have the 13r3 but I fixed my own core differential with Conductonaut. I think a large part of the reason it works so well is that other pastes just don't conduct the heat well enough to let the AW cooling system do it's job: since conductance is inversely proportional to resistance and therefore in this sense, insulation, the paste ends up holding that heat for just a bit longer rather than quickly transferring it to the heatsink and pipes/ vapor which results in the observable temperature differential. Thermal pad issues aside, it (Conductonaut) just works. Here's a temperature graph running Firestrike before and after Conductonaut on the CPU/GPU for my 13r3. (Done with HWINFO64)

Before:
https://imgur.com/GJz2EFE
After:
https://imgur.com/Q46Sd1b

That's probably why Umar has stated 'that there was no significant difference that AW observed regarding thermal pastes except with LM...' Not that it's any new info but worth noting.

 

I had a 10c drop after i also changed the thermal pads but yet still use traditional but high quality paste. The heatsink pressure definitely needs to be optimized for the best results.

 

I agree with you--not necessarily pressure but fit; I think this is why AW Dell and other manufacturers continue to use the paste that they do. All the while admitting that the results with LM are profoundly different than with other pastes.

Standard paste helps them (in a high paced production setting):

• eliminate any air gap created by incorrectly sized (thickness) pads/ heatsink face/ spring arm- deformity
• has tested and longterm data on its application
• can be quickly machine/ monkey applied and assembled by machine/ monkey
• produces a consistent result (meaning assembly)

LM:

• doesn't have a bunch of longterm data in this application
• necessitates meticulous care and attention to detail in assembly and fit and amount of LM used
• creates unknowns with regard to user serviceability because of availability of LM
• all these reasons and more would slow production and that means more cost

But the advantages of using LM ( when done right) speak for themselves, at least for the consumer.

When I re-pasted my 13 with Conductonaut, I checked for fit by: screwing down the heatsink and then lifting it again to check for the contact patch--bubbles aren't an issue with LM since it's not as viscous as normal paste and wouldn't matter anyway here. One can check by doing this(the contact patch for pad fit/ heatsink-fit (flatness to die)) and by sighting down the edge of the board to check that your electrical tape isn't above the level of the die (like sighting down a wooden board for straightness), then you know it's right.

 

So, I reported earlier that my LM repaste went well (I also opted for using Kryonaut instead of thermal pads on the associated elements around the GPU/CPU). Unfortunately, whatever I did wasn't a long term solution. Temps started climbing and last week I started seeing thermal throttling.

So, I broke my 15 R3 apart again and examined the CPU - the LM had become a hardened almost crystalline material. It appears to have oxidized and reacted to a limited degree with the copper plate in the heatsink. The copper is still intact but now has a permanent stain of gallium on the mating surface. Additionally, the surface is irregular as some of the hardened LM material has embedded.

I've read that this happens with LM; specifically the oxidation and copper reactivity. The copper reaction isn't the same as what happens with aluminum, no delamination or corruption of the structure - more like a sintering effect. The oxides were present on the CPU die along with the heatsink. I removed it using a citrus/alcohol based heatsink cleaning solution and liberal use of my elbow. I did not use any abrasives or blades and was able to return the CPU die to a mirror finish along with restoring the heatsink to a relatively flat surface.

After reading every post I could find on what is happening here, I confirmed that the Intel CPU die is indeed humped. There is a high point right at the center of the CPU and without a matching indentation in the heatsink, there will always be some sort of problem getting the cooling solution to interface properly with the die.

Of the remedies discussed here, I believe anyone looking to find a permanent fix for this problem only has three solutions:

1. If you have access to CAD tools or extreme patience, you could try to machine or lap the heatsink to account for the CPU bulge
2. Return to a conventional thermal paste that can fill the void created by the uneven mating surface
3. Use a more liberal quantity of LM and make sure you've protected surrounding components from spillage

Of the three, I opted for the last choice. My temps are back down again with differences between cores at 4C at the largest gap (Core 0 and Core 2). Torture tested it for several hours with Ghost Recon Wildlands and saw temps stabilize around 70C.

I fully expect having to do this again; absent a better seal with the LM, I'm sure the oxidation problem will return. I'm considering using a graphite or LM pad in the future. As I understand it, the graphite solution is about half as efficient as LM but has the added feature of being reusable.

 

I just finished redoing my heat sink with liquid metal following the @iunlock guide. I used the Fujipoly Extreme Pad's (17w/mk) on the areas he had listed as important, and regular Arctic pads on the less important ones. Prior to the tearing into it, my core differentials were around 15 degrees, but I would typically hit throttling about 5mins into running OCCT. When I first reassembled everything, I still had a core differential of about 10-12 Degrees with Core 2 being the highest, and Core 3 typically the lowest, but I wasn't throttling. I took it back apart, redid the LM, and did a slight modification to the CPU heat sink using a modified USB bracket from my desktop to add more pressure to the top of the heat sink. This made the results much better (as you can see in the image).

It should be noted that I'm running my overclock at 42x (or ~4.2ghz) with an undervolt of 0.075v and I also have a CCI heat sink. I have my GPU overclocked +175/+500MHz on the core and memory clocks respectively through MSI Afterburner. I just finished playing Witcher 3 for about an hour and my core temperatures had hit a high of 78C on the highest core and 74C on the lowest, with the GPU hitting 71C according to HWInfo. Let me know what you think!

https://imgur.com/jDim1oQ

 

temps are great. any picture of the hardware modification u did?

 

I'll preface this with it's a temporary modification as it's kind of "ghetto" but I plan to make a revision using a piece of aluminum and a milling machine.

I drilled out the bolt hole on the heat sink to allow it to slide over the mounting bracket on the motherboard. Then just put the clamp on the screw and tightened it down in the order of the numbers on the heat sink. It fits under the black plastic cover that goes over top of the mobo, but it does push on it a bit. Once the outer most plate is back on, it's not noticeable at all.

The clamp is from a USB-C port I had laying around from another build. I clipped off one side of the clamp with a standard set of side cutters.

https://imgur.com/a/2ad6t4l

I also added some stick on feet underneath to give it a bit of space for the fans to pull in some extra air. This seems to have helped a few degrees with the temps as well.

https://imgur.com/a/BwQXR5c

 

good mod for the weak cpu tension arm. be careful with the screw though, from your picture it looks like it might snap off any time.

 

The screw is actually in there straight, but the tapered head and the angle of the bracket make it look crooked. I've ordered some slightly longer screws with button heads for the future. I've already got an idea of the aluminum bracket I want to make and am going to have a buddy mock it up on AutoCAD for me at work. If all goes well, it should look like a factory piece.

 

Do you have better pics please ?